AUTHOR=Luo Chunlong , Wu Yang , Zhao Yi 

TITLE=SupCAM: Chromosome cluster types identification using supervised contrastive learning with category-variant augmentation and self-margin loss

JOURNAL=Frontiers in Genetics

VOLUME=Volume 14 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/genetics/articles/10.3389/fgene.2023.1109269

DOI=10.3389/fgene.2023.1109269

ISSN=1664-8021

ABSTRACT=Chromosome segmentation is a crucial analyzing task in karyotyping, a technique used in experiments to discover chromosomal abnormalities. Chromosomes often touched and occlude with each other in the images, forming various chromosome clusters. The majority of chromosome segmentation methods only worked on a single type of chromosome cluster. Therefore, the pre-task of chromosome segmentation, the identification of chromosome cluster types, requires more focus. Unfortunately, the method for this task is limited by the small-scale chromosome cluster dataset, ChrCluster, and needs the help of large-scale natural image datasets like ImageNet. We realized that semantic differences between chromosomes and natural objects should not be ignored, and thus developed a novel two-step method called SupCAM which can fully utilize ChrCluster dataset. In the first step, we pretrained the backbone network on ChrCluster following the supervised contrastive learning framework. We introduced two improvements to the model. One is called the category-variant image composition method, which augmented samples by synthesizing valid images and proper labels. The other introduced angular margin into large-scale instance contrastive loss, namely self-margin loss, to increase the intra-class consistency and decrease inter-class similarity. In the second step, we finetuned the network and obtained the final classification model. We validated the effectiveness of modules through massive ablation studies. Finally, SupCAM achieved an accuracy of 94.99% on ChrCluster dataset, which outperformed the previous method for this task. In summary, SupCAM will significantly support the chromosome cluster types identification task for better automatic chromosome segmentation.